I am attending a few sessions at Oracle OpenWorld this week and the first one is an overview of Oracle Real-Time Decisions (RTD) in e-commerce. I have reviewed RTD 3.0 before. The context for using RTD is that consumers are harder to reach, satisfy and retain – their behavior evolves more quickly, there is more competition and less loyalty, consumers offer less attention and the complexity of the problem increases constantly. Consumers’ expectations have changed too – they expect companies to treat them consistently, be relevant, use the information they have given those companies to improve their experience and make them feel special. The critical challenge is that of multiple channels – interactions cross channels and customers use more channels. Consumers might use the web to do research, call the contact center with a question, use social media to see what other people are saying and then visit a store to buy a product. It is hard for a company to ensure a seamless, effective relationship and build a deeper engagement. Over time there are many inbound and outbound moments where a company needs to optimize the next step in building the relationship with a consumer.
One of the major uses of RTD is in managing this cross-channel e-commerce deployment and many customers are focused on improving the e-commerce experience. RTD is making personalized decisions for every consumer and RTD customers have seen 200% lift in click through rates, 76% lift in sales conversion, 40% lift in margin and more as a result. One customer example was written up in a research report I did for BeyeResesarch on operational analytics and has been successful rolling out RTD for acquisition and loyalty.
Oracle RTD is described by Oracle as “Personalized business intelligence at the point of interaction” though I would say “personalized decisions at the point of interaction”. It is automated and designed to handle very large volumes and low latency – classic operational decisions. Using RTD helps you build adaptive business processes – the process remains consistent but the decision-making within that process evolves and improves constantly. RTD is a pure decision service, combining business rules and predictive analytics while also automatically applying test and learn to improve decision making and create a real-time closed loop. RTD uses predictive modeling techniques but models are created and evolved by the tool automatically rather than by a modeler. The optimization of the decision as to the best asset to be presented to the customer, and the personalization of it, are handled by RTD transaction by transaction and in very high volumes. Using RTD also generates a lot of business intelligence about the decisions being made – what works and what does not and for whom for instance.
RTD has some key characteristics:
- Can recommend any type of asset not just offers
- Recommends in context of each customer and the transaction they are performing
- Lets the user define what is best and handles multiple competing goals
- Predicts and learns continuously and automatically
- Integrates easily across multiple channels
Example uses include complete end-to-end personalization of their website, automated A/B testing, handling abandoned shopping carts and much more. Sometimes this is based on explicit customer data and sometimes it is based only on anonymous behavior as a new user or a prospect navigates a site. There are, of course, lots of decisions in e-commerce. Product recommendations, navigational content, service/product bundles, cross-sell, anonymous web visitor personalization, retention offers, email subject lines/content and much more have all been handled successfully with RTD. Example implementations exploit the RTD capabilities:
- Speed: RTD generally delivers sub-second response times such as 800+ requests with an average of <500ms on a 2 server cluster.
- Context Sensitivity: RTD can use even sparse data from an anonymous session to improve a home page so that sales increased by 9%
- Scalability: RTD in use with a 20+M customer profiles being hit in less than 10ms (using Oracle Coherence) and in another context with a catalog of more than 100,000 products
- Across lifecycle: Decisions made for anonymous visitors, identified prospects, customer service and lapsed/churned customers
- Across channels: Inbound and outbound used together to drive $100M in incremental sales from web channel and 20% lift on an email program
- Learn here, act there: Using response to RTD web content to drive outreach that improve inbound marketing results by 61%
- Multivariate testing: RTD managed 250+ test cells with intra-day changes and dramatically improved marketing velocity
- Multiple goals: Balance different objectives for existing customers and new prospects.
- Decision analysis: RTD’s reporting let one customer see a bottleneck between interest and conversion among older consumers so this could be targeted with new creative
Implementations typically go through building a foundation (hooking RTD up to channel systems, content and data sources), focusing on high-value decisions and then showing how this transforms key processes. RTD implementations can go live in a few months – a month to design, another to deploy and a third to get into production. Ongoing post-production analysis and monitoring to show the ongoing value is also key.
RTD is available directly but also tightly coupled with the Siebel CRM suite.